What ChIP-seq, ATAC-seq, and CUT&Tag Measure

Three assays, three measurement targets

ChIP-seq, ATAC-seq, and CUT&Tag all produce short DNA sequences that can be mapped to a genome, but they do not measure the same biological property. ChIP-seq enriches DNA associated with a selected protein or histone modification. ATAC-seq reports DNA regions that are accessible to an insertion enzyme. CUT&Tag uses an antibody-guided enzyme to generate fragments near a selected protein or histone modification. Their sequencing tracks may all show local read enrichment, but the meaning of that enrichment depends on how the fragments were made.

This distinction matters when interpreting regulatory regions. A signal may indicate that a region is accessible, that chromatin associated with a particular target was captured, or that an antibody-tethered enzyme generated fragments near that target. None of these measurements alone establishes that a transcription factor binds a particular DNA motif, that a gene is being transcribed, or that the measured chromatin feature causes an expression change.

What produces the sequencing signal?

In ChIP-seq, chromatin is prepared so that a target-associated fraction can be captured with an antibody. In many protocols, protein–DNA associations are crosslinked before chromatin is fragmented; other protocol choices are possible. The DNA recovered with the antibody is sequenced. A concentration of mapped fragments in a genomic region is evidence that DNA from that region was recovered with the selected target under the assay conditions. It is not a direct count of binding events. Recovery depends on the target’s presence, antibody recognition, chromatin preparation, and other assay properties.

The selected ChIP-seq target might be a transcription factor, a polymerase, or a histone modification. Those choices answer different questions. A transcription-factor ChIP-seq signal supports an association between that factor and chromatin fragments from the region. A histone-mark ChIP-seq signal supports enrichment of chromatin bearing the recognized mark. For example, enrichment for a mark commonly associated with active regulatory regions is evidence about that chromatin feature, not a measurement of the rate of transcription.

ATAC-seq does not select DNA by a target-specific antibody. Instead, the transposase inserts adapters into DNA it can access in the prepared chromatin. After sequencing, mapped fragments and their locations provide a readout of insertion events. Regions with more insertion signal are commonly interpreted as more accessible to the assay, subject to the protocol and the properties of the chromatin and enzyme. ATAC-seq does not identify which protein, if any, occupies an accessible region. Nor does every accessible region necessarily regulate a particular gene.

CUT&Tag begins with an antibody bound to a chosen target. A tethered transposase then creates fragments near the antibody-bound target, which are sequenced. Like ChIP-seq, it is target-specific in the sense that the experiment selects a protein or histone modification. Unlike ChIP-seq, it generates fragments through antibody-localized tagmentation rather than recovering an immunoprecipitated chromatin fraction. CUT&Tag can produce a strong, localized signal, but the signal remains an assay-dependent indicator of target-associated chromatin, not an exact census of target molecules.

A simple model for interpreting read enrichment

A useful way to reason about these assays is to separate the biological feature of interest from the process that turns it into reads. For a genomic region \(r\), let \(Y_r\) denote the number of mapped fragments observed there. A simplified expectation can be written as:

Here \(L\) represents sequencing depth, \(B_r\) is the relevant biological property (such as target-associated chromatin for a target-specific assay, or accessibility for ATAC-seq), and \(q_r\) summarizes assay and measurement efficiency in that region. This is a conceptual decomposition, not a complete count model. It highlights why a read count is not itself an absolute measurement of occupancy or accessibility: it reflects the biological property together with depth and assay-specific capture, fragmentation, and mapping.

For example, under the strong simplifying assumptions that depth and efficiency are the same for two samples, an expected count ratio can track a ratio in the biological property. If those assumptions do not hold, the observed ratio can change even when the biological property is unchanged. This is why raw read counts from different assays, or even different preparations of the same assay, should not be treated as directly comparable measures of molecular abundance.

Worked examples

Assumptions, diagnostics, and failure modes

Interpretation depends on the target being measured as intended and the sequence reads representing the generated fragments. For ChIP-seq and CUT&Tag, this requires that the antibody recognize the intended protein or modification sufficiently specifically and that the protocol preserve or generate target-associated fragments. For ATAC-seq, it requires that transposase insertion provide a useful proxy for accessibility in the prepared material. Across all three assays, reads must be mapped and assigned to genomic locations appropriately for a locus-level interpretation.

These assumptions are reasons to investigate unexpected patterns, not a guarantee that every signal is biologically meaningful. A target-specific signal that appears broadly where the target is not expected may reflect nonspecific recognition or other assay properties. A weak signal may reflect low target abundance, poor capture, or limited sequencing as well as absence of the biological feature. An ATAC-seq signal can be influenced by insertion preferences as well as chromatin accessibility. A region’s apparent lack of signal therefore does not prove that the target or accessibility is absent.

The appropriate diagnostics depend on the assay and protocol. At this stage, the essential interpretive check is to ask whether the observed signal is plausible for the assay’s target, whether the antibody or enzyme could generate the pattern for reasons other than the biological claim, and whether the sequencing depth and fragment locations support a stable reading of the region. Signal tracks are useful summaries, but a visible feature is not automatically evidence of a statistically established difference or a functional effect.

Choosing among the assays

The main alternative is not simply a different technology for measuring one identical quantity. The assays address overlapping but distinct questions. If the question concerns whether a specific protein or histone modification is associated with chromatin at a region, ChIP-seq or CUT&Tag is a target-specific choice. If the question concerns where DNA is accessible to transposase, ATAC-seq addresses that property without selecting one protein target. The best choice follows from the biological claim, not from which assay produces the most visually striking track.

ChIP-seq and CUT&Tag share dependence on target recognition but use different fragment-generation strategies. Neither signal is automatically a direct measure of binding strength, and comparisons between their read counts are not meaningful as absolute occupancy comparisons without an appropriate measurement framework. ATAC-seq provides a broader accessibility readout, but cannot substitute for a target-specific assay when the question is whether a particular protein or mark is present. Conversely, a target-specific assay cannot replace an accessibility assay when the question is whether DNA is open to insertion.

All three assays have limitations. They produce measurements from prepared biological material, so a bulk signal can average across cells with different states. A measured region does not, by itself, identify the gene it regulates. Assay-specific biases and experimental conditions influence observed counts. Most importantly, co-occurring signals are not causal evidence: accessibility and a histone mark at the same region do not establish that either feature produced a gene-expression change.

Common mistake

Wrong reasoning: “ATAC-seq has a strong signal at this region, so the transcription factor must bind there and activate the nearby gene.” This conclusion moves from accessibility to a specific protein’s binding and then to a causal expression effect—three different claims.

Correction: ATAC-seq supports a conclusion about transposase-accessible DNA in the assayed material. A target-specific assay is needed to investigate association with a particular protein, and separate evidence is needed to establish transcriptional consequences or regulatory causality. Even agreement among assays strengthens a biological interpretation without making those claims automatic.